"Tokenmaxxing" is the latest corporate trend taking over the tech world, but is it actually measuring developer productivity—or is it just burning a monumental hole in company finances?
In this episode of Internet of Bugs, Carl breaks down the origins of tokenmaxxing, from Jensen Huang’s podcast comments to prestigious VC letters pushing AI-driven developer metrics. We explore the terrible history of trying to measure programmer productivity (like counting lines of code), how Goodhart's law ruins these systems, and why massive tech companies might actually be incentivizing this slop just to train their own internal AI models.
Furthermore, we look at the financial reality of the AI bubble, upcoming tech IPOs, and why shipping dead code or hiding bugs behind feature flags can cause catastrophic real-world losses (like the 2012 Knight Capital Group meltdown).
💡 SUPPORT THE CHANNEL ON PATREON:
Join our Friday Patreon Q&A Zoom calls to talk about upcoming videos, tech trends, and industry superstitions sooner!
👉 / internetofbugs
- 00:00 What is "Tokenmaxxing"?
- 00:55 Where the Trend Started (Nvidia & VCs)
- 02:03 The Nightmare of Corporate Token Leaderboards
- 02:24 How Do We Actually Measure Good Code?
- 03:52 Lines of Code, Story Points, and Failed Metrics
- 05:01 Goodhart's Law & Gaming the System
- 05:39 Why Big Tech Wants You Burning Tokens
- 06:49 The Superstitions of Software Engineering (LeetCode & Agile)
- 07:55 The Danger of Dead Code & The $440M Bug
- 09:17 The Real Solution
- 10:47 The AI Bubble's "Groundhog Day" IPO Forecast
#Tokenmaxxing #SoftwareEngineering #AIBubble #DeveloperProductivity #TechTrends #Programming #TechIndustry #InternetOfBugs
// Sources
// Origins
https://the-decoder.com/nvidia-ceo-je...
https://www.nytimes.com/2026/03/20/te...
https://a16z.com/there-are-only-two-p...
https://finance.yahoo.com/news/a16z-u...
https://www.notboring.co/p/a16z-the-p...
https://en.wikipedia.org/wiki/Andrees...
// What companies are doing
/ metas_internal_leaderboard_ranks_employees...
https://fortune.com/2026/04/09/meta-k...
https://www.bbc.com/news/articles/cvg...
https://www.firstround.com/ai/shopify
https://gaiinsights.com/blog/the-ceo-...
https://help.shopify.com/en/manual/sh...
https://www.forrester.com/blogs/what-...
https://www.businessinsider.com/amazo...
https://aws.amazon.com/blogs/machine-...
https://aws.amazon.com/blogs/machine-...
// Programming History Mentioned
https://maseconomics.com/goodharts-la...
https://dl.acm.org/doi/10.1145/362929...
https://junit.org/
https://www.henricodolfing.ch/en/case...
https://blog.statsig.com/how-to-lose-...
// AI limitations
https://epoch.ai/blog/will-we-run-out...
// Tokenmaxxing backlash
https://www.tomshardware.com/tech-ind...
https://news.ycombinator.com/item?id=...
https://www.businessinsider.com/uber-...
// Financial irregularities (just the most recent tip of the massive iceberg)
https://www.wheresyoured.at/anthropic...
https://www.thestreet.com/latest-news...
https://www.datacenterdynamics.com/en...
https://techcrunch.com/2026/05/20/ant...
Java a évolué.
Votre code peut aussi.
Une collection de snippets Java modernes. Chaque ancien pattern Java à côté de son remplacement moderne et propre — côte à côte.
marché sur macos
Zorin OS est une distribution Linux basée sur Ubuntu. Elle utilise par défaut un environnement de bureau GNOME.
Le bureau est adapté afin d'aider les utilisateurs à passer facilement de Windows, macOS ou ChromeOS à un environnement Linux. Zorin OS est fourni avec Wine et PlayOnLinux en natif pour l'exécution de logiciels Windows
Photogimp fait ressembler Gimp a photoshop et facilite la transition.
Be sure to use AI when making
your next, I don’t know, meal plan,
for example. Definitely do not call
your friend who loves to cook and ask her
for her favorite recipes or tips or ways
to save time making meals,
because you will end
up talking for longer than you had hoped,
hearing, perhaps, about her father’s cancer
diagnosis or how lonely she’s been or even
what she’s planted in her spring
garden and then lost with the early frost.
...
Automated tests are important. Without them, programmers waste a huge amount of time manually checking and fixing their code.
Unfortunately, many automated tests also waste a huge amount of time. The easy, obvious way to write tests is to make broad tests that are automated versions of manual tests. But they’re flaky and slow.
Folks in the know use mocks and spies (I say “mocks” for short in this article) to write isolated interaction-based tests. Their tests are reliable and fast, but they tend to “lock in” implementation, making refactoring difficult, and they have to be supplemented with broad tests. It’s also easy to make poor-quality tests that are hard to read, or end up only testing themselves.
Bad tests are a sign of bad design, so some people use techniques such as Hexagonal Architecture and functional core, imperative shell to separate logic from infrastructure. (Infrastructure is code that involves external systems or state.) It fixes the problem... for logic. But infrastructure is often left untested, and it requires architectural changes that are out of reach for people with existing code.
This pattern language1 describes a fourth option. It avoids all the above problems: it doesn’t use broad tests, doesn’t use mocks, doesn’t ignore infrastructure, and doesn’t require architectural changes. It has the speed, reliability, and maintainability of unit tests and the power of broad tests. But it’s not without tradeoffs of its own.
1The structure of this article was inspired by Ward Cunningham’s CHECKS Pattern Language of Information Integrity, which is a model of clarity and usefulness.
The patterns combine sociable, state-based tests with a novel infrastructure technique called “Nullables.” At first glance, Nullables look like test doubles, but they're actually production code with an “off” switch. And that’s the tradeoff: do you want that in your production code? Your answer determines whether this pattern language is for you.
The rest of the article goes into detail. Don’t be intimidated by its size. It’s broken up into bite-sized pieces with lots of code examples.
- TTY
- tmux
- kmscon
- syncthing
- nmtui
- neovim
- Size font
- more than 16 colors
- Alternate typeface
- rudimentary mous support
idée de format de kata : "Skill issue"
Basée sur l'idée de "mauvaise foi dans le code, bonne foi dans les tests". (poke @romeu )
Driver joue le rôle d'un agent IA de mauvaise foi.
Driver dispose de quelques fichiers skill.md initialement un peu ambigües.
Driver les interprète de la pire façon.
Ensemble du groupe doit changer ses instructions et/ou les skills pour que Driver fait ce qui est attendu.
Le kata se fait sans L'utilisation de vrai LLM (indeed completion AI)
